tw-research-proposal-diamond
Pass
Audited by Gen Agent Trust Hub on May 18, 2026
Risk Level: SAFECOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to install the 'docx' package from the official npm registry using 'npm install -g docx'. It also executes Node.js scripts and a local Python validation script at '/mnt/skills/public/docx/scripts/office/validate.py'. These actions are functional requirements for generating formatted academic documents.\n- [REMOTE_CODE_EXECUTION]: The agent is directed to generate and run Node.js code based on templates provided in 'references/docx-output-guide.md' to transform Markdown content into a Word document. This is a standard method for document automation within the agent environment.\n- [PROMPT_INJECTION]: The skill presents an indirect prompt injection surface by processing external data from web searches and user responses. \n
- Ingestion points: User-provided research interests and live results from the 'web_search' tool across all four phases.\n
- Boundary markers: Absent; there are no specific instructions to the agent to treat external search results as untrusted data.\n
- Capability inventory: 'web_search', shell command execution (Node.js/Python), and file system writes to the user data directory.\n
- Sanitization: Absent; the instructions do not specify validation or escaping for the content retrieved during research.
Audit Metadata